[PATCH] Remove spurious BUG from posix-cpu-timers

[PATCH] Remove spurious BUG from posix-cpu-timers

[Matthew Wilcox]

PA-RISC does not necessarily have interrupts disabled while timers are
running.  This doesn't appear to be a requirement, but posix-cpu-timers
seems to think it is.

Signed-off-by: Matthew Wilcox <matthew@wil.cx>

diff -urpNX dontdiff linus-2.6/kernel/posix-cpu-timers.c parisc-2.6/kernel/posix-cpu-timers.c
--- linus-2.6/kernel/posix-cpu-timers.c	2005-03-18 05:55:27.000000000 -0700
+++ parisc-2.6/kernel/posix-cpu-timers.c	2005-03-18 13:52:53.341505966 -0700
@@ -1242,8 +1242,6 @@ void run_posix_cpu_timers(struct task_st
 	LIST_HEAD(firing);
 	struct k_itimer *timer, *next;
 
-	BUG_ON(!irqs_disabled());
-
 #define UNEXPIRED(clock) \
 		(tsk->it_##clock##_expires == 0 || \
 		 cputime_lt(clock##_ticks(tsk), tsk->it_##clock##_expires))

-- 
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain

Re: [PATCH] Remove spurious BUG from posix-cpu-timers

[Andrew Morton]

Matthew Wilcox <matthew@wil.cx> wrote:
>
> 
> PA-RISC does not necessarily have interrupts disabled while timers are
> running.  This doesn't appear to be a requirement, but posix-cpu-timers
> seems to think it is.
> 

But a bit later on we do

        spin_lock(&tsk->sighand->siglock);

rather than spin_lock_irqsave()?

Re: [PATCH] Remove spurious BUG from posix-cpu-timers

[Matthew Wilcox]

On Fri, Mar 18, 2005 at 08:53:34PM -0800, Andrew Morton wrote:
> Matthew Wilcox <matthew@wil.cx> wrote:
> >
> > 
> > PA-RISC does not necessarily have interrupts disabled while timers are
> > running.  This doesn't appear to be a requirement, but posix-cpu-timers
> > seems to think it is.
> > 
> 
> But a bit later on we do
> 
>         spin_lock(&tsk->sighand->siglock);
> 
> rather than spin_lock_irqsave()?

We're in interrupt context ... we're ok, unless another interrupt handler
tries to take the same spinlock.  Maybe the posix cpu timer code should
be using _irqsave anyway?

-- 
"Next the statesmen will invent cheap lies, putting the blame upon 
the nation that is attacked, and every man will be glad of those
conscience-soothing falsities, and will diligently study them, and refuse
to examine any refutations of them; and thus he will by and by convince 
himself that the war is just, and will thank God for the better sleep 
he enjoys after this process of grotesque self-deception." -- Mark Twain

Re: [PATCH] Remove spurious BUG from posix-cpu-timers

[Roland McGrath]

I put those BUG_ON's in places where I was using spin_lock on a lock that
requires irqs disabled.  Those places are called in contexts that already
require irqs disabled for some other reason, so I didn't use
irqsave/irqrestore there to skip the unnecessary overhead.  (Of course it's
the same the BUG_ON's have, but those can be compiled out.)  I really don't
recall at the moment if there was a deeper reason, i.e. that the code itself in
some other way relies on interrupts being disabled for its proper operation.
But I don't see any reason to change it.


Thanks,
Roland